Invited keynote talk given to the Malaysia International Conference on Languages, Literatures, and

Cultures (MICOLLAC), May 2007.

Interpreting Experience:
Clues from the Study of Language Acquisition
William O’Grady
University of Hawai‘i

1. The frequency factor

All three of the disciplines represented at this conference are deeply and
intimately concerned with experience: language allows us to express it, literature
helps us interpret it, and culture offers ways to organize and institutionalize it. This
in turn leads to the following fundamental question, which also happens to be a
central concern of cognitive science: how does the mind make sense of experience,
not only in the formative early years, but also throughout life?
It’s not clear that this question can or should be approached in the same way in
the study of language as it is in the study of literature or the study of culture.
However, it is perhaps not unrealistic to hope for parallels in the way in which the
question might be investigated. At the very least, the study of how the mind
responds to experience in one domain may yield findings that can be contrasted and
compared with those obtained from the investigation of other domains. With that in
mind, I propose to explore this question from a linguistic point of view, offering
suggestions that will, presumably, be of greatest interest to those working in the
fields of linguistics and applied linguistics, but may nonetheless raise questions (or
objections) of interest to scholars in literary and cultural studies.
A very central question in the study of language and its relationship to
experience has to do with the role of frequency in shaping and explaining the
character of linguistic phenomena. Put simply, to what extent does the frequency
with which particular forms and patterns are encountered affect their properties,
including the rate at which they are acquired?
For several decades, the linguistic mainstream has been deeply skeptical about
the relevance of frequency to explanatory work on language. The roots of this
skepticism can be traced to Noam Chomsky’s seminal work, Syntactic Structures
(1957:16), in which he observed that ‘…one’s ability to produce and recognize
grammatical utterances is not based on notions of statistical approximation and the
like.’
More recently, however, it is has become evident that frequency has quite
fundamental effects on various aspects of language. Here are a few examples:
•Frequent forms, by virtue of being used and heard so often, tend to maintain any
irregularity that they have. For example, the most enduring irregular past tense
forms in English occur with high-frequency verbs such as be, go, come, and so
forth (e.g., Pinker 1999:125). In contrast, less frequent verbs such as heave and
thrive have lost their irregular past tense forms (hove and throve, respectively).
 2

•Frequent forms tend to develop special ‘shortened’ pronunciations. For instance

the final ‘t’ or ‘d’ is more likely to be deleted in high-frequency words such as
just (‘He jes left’), went (‘I wen there yesterday’) and don’t (‘don do it’) than in
low-frequency words such as innocent, interest, or attract.1 Similarly, an
unstressed vowel is more likely to be deleted in the middle of a high-frequency
word such as every than in a low-frequency word such as summery. Mid-
frequency words such as memory and family permit variation. (Bybee &
McClelland 2005:390).
•Almost all of the full forms currently involved in contraction in English (don’t, I’m,
wanna, they’re, there’s, etc.) come from among the 150 most frequent words in
the language (Krug 1998:294).
•The internal structure of affixed words reflects the relative frequency of the word
compared to that of its stem. Because adequate is more frequent than the affixed
form inadequate, its presence in the derived word is relatively salient, leading to a
high native speaker rating for structural complexity. In contrast, because audible is
less frequent (and therefore less salient) than inaudible, the affixed form receives a
low rating for structural complexity. (Hay 2003)
•Words such as swiftly (which is more frequent that swift and therefore has a
relatively weak internal morpheme boundary) are pronounced with less of a stem-
final [t] than words such as softly, which is less frequent than soft and therefore
has a relatively strong internal morpheme boundary. (Hay 2003)
•More frequently occurring patterns are processed more quickly and more
successfully than infrequently occurring patterns (e.g., active sentences vs.
passives; see Townsend & Bever 2001 and Dick et al. 2001). 2
•There is a strong inverse correlation between length and frequency: more frequent
forms tend to be shorter (e.g., Haspelmath 2006:43 and the references cited there).
For additional examples and discussion, see Bybee & McClelland (2005), Bybee &
Hopper (2001), Ferreira (2003), Croft (2003:110ff), and Haspelmath (to appear).
There are also various well-established examples of frequency effects in the case
of language acquisition.
•Frequency, by itself or interacting with factors such as phonetic salience, affects the
developmental order of ‘grammatical morphemes’ (Brown 1973, Goldschneider
& deKeyser 2005). For example, plural -s (as in two dogs), which is acquired
before the homophonous verbal -s suffix (as in He works hard), occurs more
frequently overall and more frequently in the salient utterance-final position (Li,
Leonard & Swanson 1999).
1
For more on this see, Jurafsky et al.. (2000), who note that that likelihood of a final t/d being
deleted increases with the word’s frequency, regardless of whether it is a content word or a function
word.
2
Ferreira reports (p. 194) that a processing model with two ‘fast and frugal heuristics’ (the NVN
strategy and a semantic association heuristic—animates tend to be agents; inanimates tend to be
themes) can account for a large amount of data involving the relative processing difficulty of
canonical and non-canonical sentences.
 3

Relative frequency of two -s suffixes in parental speech sample

(Li et al, p. 537)

all positions combined end of the sentence

plural -s 285 148
verbal -s 55 9

•Children are relatively quick at figuring out the right past tense form for frequently
heard irregular verbs like go and see, but that they often take much longer to
master the irregular past tense form of less common verbs such as sink or win.
Maratsos (2000) estimates that it may take several hundred exposures to the
correct form of a verb before all over-regularizations are eliminated.
•Children’s early production of direct objects reflects properties of the input.
Thus children tend to use a direct object with verbs such as get and like, which
are almost always used transitively by their parents, but often drop direct
objects with verbs such as eat, which are frequently used intransitively in
parental speech (Ingham 1993/1994, Theakston, Lieven, Pine & Rowland 2001).

But to what extent are such examples typical of language acquisition in general?
Absolutely typical, according to some. For instance, Ellis (2006:1) has gone so far
as to suggest that language acquisition is ‘contingency learning, that is the gathering
of information about the relative frequency of form-function mappings.’
At first glance, this seems unlikely since there appear to be obvious examples of
phenomena that are manifested with a high degree of frequency in the input yet very
difficult to acquire. The English articles the and a are a case in point. For the sake of
exposition, let us focus just on the.
The is the most frequent word in the English language, yet it is a relatively late
acquisition. This is obviously the case for second language learners, who find the
English article system notoriously difficult and very often fail to attain native-like
mastery of its use even after many years of instruction and practice. It is even true,
although obviously to a much lesser extent, for child first language learners: Brown
(1973:358) reports that use of the articles the and a attains 90% accuracy
considerably later than several less frequent morphemes, including the progressive
marker -ing, the plural suffix -s, the possessive morpheme -’s, and irregular past
tense. How can we reconcile such facts with the idea that frequency has a major role
in shaping the language acquisition process?
There is, I think, a potentially satisfying answer to this question that comes from
one of the few real points of consensus in the field of linguistics, namely that
language offers its users a way to map meaning onto form and vice versa. That is,
we use words and sentences to express meaning to others, and we extract meaning
from the words and sentences that others direct toward us.
 4

If this is right, then the language acquisition process is devoted in large part to
uncovering the particular mapping between form and meaning that is employed by
the language to which the learner (child or adult) is exposed. This in turn suggests
that to the extent that frequency is important to language acquisition, what counts is
not how many times learners hear a particular form—it is how many times they
encounter particular mappings between form and meaning.
Returning now to the problem of the acquisition of English articles, the crucial
question isn’t whether language learners hear the more than any other word, it is
how often they encounter instances of that particular form being used with an
unequivocal meaning. There is reason to believe that this happens far less frequently
than one might suspect. Two observations are crucial.
First, as observed by Ellis (2006b:171) grammatical functors in general can be
identified based on ‘bottom-up’ (purely acoustic) evidence as little as half the time
by adult native speakers (Herron & Bates 1997).3
Second, in the case of determiners, the meaning side of the form-meaning
mapping is quite opaque and difficult to discern. As Ionin, Zubizaretta &
Maldonado (2007) observe, at least two factors are potentially relevant to article
systems in languages of the world:
•Definiteness: whether prior discourse, world knowledge, and so on establish the
existence and uniqueness of the referent for the hearer as well the speaker
•Specificity: whether the speaker intends to refer to a particular referent, which has
some property that makes it noteworthy from his/her point of view.
It turns out that the English article system is used primarily to encode definiteness,
whereas the system employed in Samoan (for instance) is focused on specificity.
This creates two sorts of problems for language learners. First, the contrast
between definiteness on the one hand and specificity on the other is a subtle one,
since it presupposes access to the attitude and intentions of the speaker and the
hearer—something which is often not fully available, of course. Matters are further
complicated by the fact that definites are often specific, as in the following example
from Ionin et al.:

(1) [+definite, +specific]

I want to talk to the owner of this store—she is my neighbor, and I have an
urgent message for her.

Even assuming that a learner listening in on this conversation could infer the state of
mind of both speaker and hearer, it would not be evident why the is used. Is it
3
In the Herron & Bates studies, subjects had to identify semi-homophonous function words and
content words (the pronoun I and the noun eye, the auxiliary verb will and the noun will, and so
on) that had been spliced into various contexts—some neutral, some appropriate for just the
content word, and some appropriate for just the function word.
 5

because of definiteness—the speaker is referring to someone whose existence and

uniqueness are established by general world knowledge (typically a store has a
unique owner)? Or is it because of specificity—the speaker intends to refer to
someone with a noteworthy property (she is a neighbor for whom she has an urgent
message)? There is basically no way to tell, which means there is not an unequivocal
mapping between the form the and its meaning in this case and in many others, as
Ionin et al. note (p. 20).
That is not to say that the function of the definite article is never clear. Here is
one such context, also from Ionin et al.

(2) [+definite, -specific]

I want to talk to the owner of this store, whoever that is. I am going to
complain about the quality of the produce!

In this case, the speaker’s comment to the effect that she does not have in mind a
specific referent allows the learner to infer that the is licensed by the existence of a
unique (if unknown) referent—i.e., by definiteness. Crucially, however, it is unlikely
that contexts in which the meaning of the is so unequivocally clear will be very
numerous. As Ionin et al. note (p. 20):

Given the subtlety of the discourse triggers related to speaker and hearer
knowledge, generalizing from them is likely to be a fairly long and difficult
process.

In the final analysis, then, although the form the may indeed occur very
frequently, mappings between the form and its meaning are far, far less frequent.
Indeed, they may even be downright infrequent. The late acquisition of the English
article system therefore need not count against the idea that frequency has a crucial
role to play in shaping language acquisition.

2. Why frequency is not enough

Thus far, I have been defending the idea that there are genuine frequency effects
in language acquisition. I have even tried to make more precise what type of
frequency might be of greatest relevance to language learners, suggesting that we
need to focus on the rates of occurrence of particular form–meaning mappings,
rather than on just the occurrence of particular forms.
Nonetheless, I remain deeply skeptical about whether frequency effects will
really turn out to be central to our understanding of language acquisition and use.
Recall that the general problem we are considering has to do with the interpretation
of experience—in particular, the question of the particular properties of experience
that the mind cares about the most. Experience does not leave its imprint directly on
 6

the mind—it passes through a processor that filters out certain things and focuses
on others. For example, phonological processing focuses on just those features of
sound that are linguistically relevant, setting to the side a wide range of extraneous
factors ranging from background noise, to a foreign accent, to symptoms of a cold.
So the question really comes down to this: what type of processors do we have?
Are they sensitive primarily to how often particular types of experience (e.g.,
particular structural patterns) are encountered, or are they guided by other
considerations? My own recent work (e.g., O’Grady 2005) places a great deal of
emphasis on this particular question since I take as my starting point the idea that
the processor lies at the heart of the human language faculty and that a proper
understanding of its functioning can contribute in major ways to understanding how
language works and how it is acquired.
The standard view within linguistics is that the processor is at the service of a
(largely innate system of) grammatical rules that determine which sentences of a
language are well formed, what they can and cannot mean, and so on. My view is
quite different. I have grave doubts about whether grammatical rules exist, or at least
whether they exist in the form in which they are traditionally conceived. And I
believe that the primary goal of the processor is simply to minimize the burden on
working memory—the pool of resources that not only holds representations but
also supports operations on those representations (e.g., structure building,
interpretation, and so forth; see O’Grady 2005:6).
Let’s consider some examples, beginning with the familiar contrast between
reflexive pronouns and plain pronouns.

2.1 Pronouns
At some point, every language learner (and every language teacher) has to
confront the distinction between reflexive pronouns such as himself and herself on
the one hand and plain pronouns such as him and her on the other. Both types of
pronoun introduce a referential dependency—roughly speaking, they require the
search for an ‘antecedent’ that can identify their referent. But different sorts of
referential dependencies are evidently involved since the two types of pronouns are
not in general interchangeable. (I’ll use subscripts to indicate a pronoun-antecedent
pair.)

(3)a. Jerryi admires himselfi.

b. Jerryi admires himj.

Notice how, in this very simply example, himself necessarily refers to Jerry, while
him necessary refers to someone other than Jerry, presumably someone mentioned
in the preceding context. A similar contrast, at least in monoclausal sentences exists
in Malay (Cole & Hermon 2005:633-35).
 7

(4) a. Johni mengajar dirinya sendirii.

John teach self-3SG alone
‘Johni is teaching himselfi.’

b. Johni mengajar diaj.

John teach 3SG
‘Johni is teaching him/herj.’

Returning now to English, for which we have of course more information, there
is good reason to believe that plain pronouns occur far more frequently than their
reflexive counterparts. One indication of this comes from a search of maternal
speech to Adam, Eve and Sarah that I conducted in the CHILDES data base. The
search turned up 17 instances of himself, compared to 487 instances of him. There
was an even sharper contrast between themselves, which occurred just once, and
them, which appeared 717 times. All other things being equal, then, plain pronouns
should be mastered before reflexives.
Interestingly, however, development runs in the opposite direction: the literature
generally reports that whereas reflexive pronouns are mastered by age 3 or so,
children continue to have significant difficulty with plain pronouns until around age
6. (See O’Grady 1997:226ff for a review.4 ) Thus children who are asked whether
‘Mama Bear is touching her’ in a picture such as the one below will often answer
‘yes,’ treating her as if it meant herself.

Figure 1: ‘Is Mama Bear touching her?’

Similar errors, although less well documented, have been reported for second
language acquisition (Finer & Broselow 1986, Cook 1990, White 1998). Crucially,
however, the reverse error (interpretation of a reflexive pronoun as if it were a plain
pronoun) appears to occur only rarely, if at all, in either first or second language
acquisition.
If frequency is not responsible for the development of the contrast between plain
and reflexive pronouns, then what is? I explore one possibility in Syntactic

4
More recent work suggests that children may be less prone to err in the interpretation of plain
pronominals than was previously thought, although this type of pronoun still creates more
problems for language learners than does its reflexive counterpart (Conroy et al. 2006).
 8

Carpentry (O’Grady 2005:28ff). The key idea is that reflexive pronouns differ
from plain pronouns in being immediately interpretable.
To see how this works, let us assume that sentence are formed and interpreted
one word at a time from left to right, with all dependencies—including the referential
dependencies introduced by pronouns—being resolved at the first opportunity.
Then a sentence such as John overestimates Bob will be formed as follows. (I
indicate that a nominal has satisfied one of a verb’s argument dependencies by
copying its index into the grid of the verb.)

(5) a. Combination of the verb with its first argument, the NP John; the index of
the nominal is copied into the verb’s argument grid:
[Johni overestimates].
<Ni N>

b. Combination of the verb with its second argument, the NP Bob, whose index
is then copied into the verb’s argument grid:
[Johni [overestimates Bobj]].
<Ni Nj >

Matters work essentially the same way when the sentence contains a reflexive
pronoun, whose interpretation requires resolution of the referential dependency that
it introduces. As shown below, the referential dependency can be instantly resolved
by the index of John in the grid of the verb with which the pronoun combines.

(6) a. Combination of the verb with its first argument, the NP John; the index of
the nominal is copied into the verb’s argument grid:
[Johni overestimates].
<Ni N>

b. Combination of the verb with its second argument, the reflexive pronoun;
resolution of the referential dependency:
[Johni [overestimates himselfx]].
<Ni Nx >
↓ <resolution of the referential dependency>
i

In contrast, a plain pronoun in the same position must be linked to a referent

introduced at an earlier point in the discourse.
 9

(7) [Johni [overestimates himx]].

<Ni Nx >
|
?<------------------------

Independent experimental work by Piñango, Burkhardt, Brun & Avrutin (2001) and
Sekerina, Stomswold & Hestvik (2004) confirms the increased processing difficulty
associated with locating and identifying such referents. In Sekerina et al.’s eye
movement study, for instance, children manifested a strong preference (as
manifested in looks toward potential referents) for an antecedent in the same
sentence for both the plain pronoun and the reflexive pronoun in patterns such as
The boy placed the box behind him/himself.
If all of this is on the right track, then there is a simple but far-reaching lesson to
draw: computational efficiency (the burden on the processor) is a better predictor of
development than just frequency when the two are in conflict.

2.2 Relative clauses

For many years, relative clauses have provided revealing opportunities to further
our understanding of the language acquisition process. Particularly instructive is the
contrast between subject relative clauses such as (8a) and direct object relatives such
as (8b).

(8) a. Subject relative clause:

the student [who saw the woman]

b. Direct object relative clause:

the woman [who the student saw]

Based on a study of the speech of four mothers to their children, Diessel (2004)
reports that direct object relatives are more frequent than subject relatives.

Subject relatives: 35.6%

Direct object relatives: 56.8%
Other: 7.6%

From the point of view of frequency then, we would expect direct object relatives to
be easier than subject relatives. But this is not what we find: rather, study after study
has shown that subject relatives are mastered before direct object relatives, in the
case of both first language acquisition (see O’Grady 1997: 226ff for a review) and
second language acquisition (see Hamilton 1994 for a review). In addition, it is
known that patients suffering from agrammatic aphasia find it far easier to
understand subject relative clauses than direct object relatives (e.g., Grodzinsky,
1989).
 10

Moreover, we know from experimental work by Roth (1984) that even with
explicit training, three and four year olds maintain a strong preference for subject
relatives over direct object relatives. As Caplan & Waters (2002:71) observe, ‘no
amount of practice … seems to be able to make object-relative sentences as easy to
process as subject-relative sentences.’
An additional fact, familiar to many of you, is also worth mentioning here: as
documented by Keenan & Comrie (1977), subject relatives are typologically less
marked than their direct object counteparts. Thus there are languages that permit
subject relatives but not direct object relatives (but not vice versa). Indeed, as Wong
(2002:64) observes, Malay is such a language. Thus the direct object relative in (9),
which is more or less a direct translation of its English counterpart, is unacceptable
to many speakers of standard Malay.

(9) *Doktor [yang John menolak _] ialah Dr.Tan.

doctor that John push is Dr. Tan
‘The doctor [who John pushed _] is Dr. Tan.’

Instead, as Wong notes, the verb must be passivized, resulting in a subject relative
clause.

(10) Doktor [yang _ ditolak oleh John] ialah Dr.Tan.

doctor that PASS-push by John is Dr. Tan
‘The doctor [who _ was pushed by John] is Dr. Tan.’

Why should there be such a far-ranging asymmetry between subject and direct
object relative clauses? The answer, I believe, lies in yet another long-standing
finding: subject relatives are easier to process than their direct object counterparts
(e.g., Wanner and Maratsos, 1978; Gibson, 1998; Caplan and Waters 2002),
apparently because of a difference in the demands made on working memory in the
course of forming the two types of structure. As illustrated below, a typical relative
clause begins with a relative pronoun (e.g., who) that must be associated with a
position in the verb’s argument grid—the first position in the case of the subject
relative and the second position in the case of the direct object relative.

(11)a. Subject relative clause:

the student [who i saw the womanj]
<N i N>

b. Direct object relative clause:

the student [who j the womani saw]
<Ni N j >
 11

A useful way to think about the relationship between the relative pronoun and the
corresponding position in the verb’s argument grid is to assume that wh words
introduce a ‘dependency’ that is resolved with the help of a previously open
position in a verb’s argument grid (see O’Grady 2005:113ff for details).
Like referential dependencies, wh dependencies are easiest when resolved at the
first opportunity. Crucially, as the contrast in (11) illustrates, that opportunity arises
sooner in the case of subject relative clauses, in which the relative pronoun occurs
next to the verb, than in direct object relatives, where the relative pronoun is
separated from the verb by the intervening subject. This in turn suggests that object
relative clauses may make comparatively greater demands on working memory.
Evidence from processing studies offers striking support for this idea.
In a classic experiment, Wanner & Maratsos (1978) displayed sentences
containing relative clauses on a screen one word a time, temporarily interrupting the
presentation after the fourth word with a list of five names.

(12)a. Subject relative clause:

The witch [who despised sorcerers] frightened little children.
⇑
interruption point [Bob, Henry, Mary, Jean, Tom]

b. Direct object relative clause:

The witch [who sorcerers despised] frightened little children.
⇑
interruption point [Bob, Henry, Mary, Jean, Tom]

Recall of names and comprehension of the relative clauses were tested at the end of
the sentence, and both were significantly poorer for the direct object relative clauses
than for the subject relatives.
This is just what one would expect if establishing the link between the wh word
and the verb taxes working memory. Because subject relative pronouns occur
adjacent to the verb, the relationship can be established immediately, before the
interruption point in (12a).

(13) Subject relative clause:

The witch [who despised sorcerers] frightened little children.
↑ ⇑
wh dependency interruption point
is resolved here

In contrast, the object relative pronoun in (12b) is more distant from the verb and
establishment of the link between the two is impeded by the intervening interruption,
with negative consequences both for recall of the intruding names and for
interpretation of the sentence.
 12

(14) Direct object relative clause:

The witch [who sorcerers despised] frightened little children.
⇑ ↑
interruption point wh dependency
is resolved here

Here again, then, we see an instantiation of the general point also exemplified in
the case of pronoun interpretation. Processing difficulty trumps frequency in
determining the course of development.

2.3 The positioning of prepositional phrases

It is well known that English has a strong preference for a fixed canonical word
order in its basic clauses. A very basic example of this involves the subject – verb –
direct object – prepositional phrase order found in sentences such as (15).

(15) Mary placed [the book] [on the shelf]

subject–verb–direct object–prepositional phrase

Although this order is of course by far the most frequent one for sentences
containing a subject, a direct object and a prepositional phrase, this is not enough to
ensure its correct interpretation, at least in the case of second language acquisition.5
Suggestive data in this regard comes from a study that two colleagues and I did of
the ability of college-level Japanese- and Korean-speaking learners of English to
respond to simple commands such as the following (O’Grady, Yamashita & Lee
2005a,b):

(16) Basic instrumental pattern

Tap the pencil with the chalk.

(17) Fronted instrumental pattern

With the chalk tap the pencil.

(18) Basic locative pattern

Put the chalk on the pencil.

There were two key findings:

•Learners did far better on the basic locative pattern than on the basic instrumental
pattern, even though both employ exactly the same canonical word order.

5
Comparable problems have been reported for agrammatics (O’Grady & Lee 2005).
 13

•Learners did far better on the fronted instrumental pattern than on the basic
instrumental pattern, even though the fronted pattern employs a rare non-canonical
word order.
The explanation that we proposed for these results involved processing: sentences
are easier to process if their word order aligns with the structure of the
corresponding event (the Isomorphic Mapping Hypothesis). Consider in this regard
the contrast between the two instrumental patterns—Tap the pencil with the chalk
versus With the chalk, tap the pencil. In order to comply with either version of this
request, one first has to pick up the chalk and then use it to touch the pencil—a
sequence of events that better matches the non-canonical word order in the second
pattern, where the chalk is mentioned first, than the canonical word order in the first
sentence.

Figure 2: ‘With the chalk, tap the pencil.’

(vs: ‘Tap the pencil with the chalk.’)

No such anomaly arises in the case of a locative pattern such as Put the chalk on
the pencil. In order to carry out this request, one first picks up the chalk (which is
also mentioned first) and then places it on the pencil—the sequence of events
matches the canonical word order.
Once again then, we see a conflict between frequency on the one hand
(represented by the basic instrumental pattern) and ease of processing (instantiated
by the fronted instrumental pattern, whose order aligns with the structure of the
event). And once again we see that ease of processing wins out.6

6
Ferreira (2003:190) also observes that the base frequency of a ‘global syntactic form’ is not what
determines how accurately people comprehend a sentence, noting that subject clefts (e.g., It was
John that saw Mary) are no harder to understand than regular active sentences.
 14

3. Conclusion
I believe that the phenomena we have been considering are typical and therefore
instructive in a very general way. There are indeed frequency effects in language
acquisition, as we saw in the first part of this paper. However, consideration of a
broader range of examples suggests that such effects are quite weak and that they
give way to more compelling demands of the processor. In particular, what really
matters is not how frequent a particular pattern is, but rather how much burden it
places on the processor. Patterns that are associated with a relatively high processing
cost emerge more slowly in the course of language acquisition—hence the relative
difficulty and delayed mastery of plain pronouns compared to reflexives, of object
relatives compared to subject relatives, and of word orders that align with the
structure of the event compared to those that do not.
The general lesson here is in fact a familiar one: experience does not leave its
imprint directly on the mind. Rather, as I suggested earlier, its effects and impact are
shaped by the mental mechanisms that perceive and analyze it. In the case of
language, the most important of these mechanisms is a simple processor whose
primary concern is minimizing the burden on working memory, as we have seen.
Understanding this simple fact can provide valuable insights not only into how
language works, but also into why it is acquired in the particular way that it is and
perhaps even how, in the case of second language learning, we might go about
developing pedagogical materials that target phenomena that are particularly
difficult. (For remarks along different lines on the relevance of frequency
information to grammar teaching, see Biber & Reppen 2002.)
In the longer term, it may even be feasible to pursue in more detail the
possibility, raised at the outset, of uncovering meaningful cross-disciplinary parallels
in the interpretation of experience. I don’t think that language is alone in relegating
the frequency of particular types of experience to a peripheral role. The grand
themes of literature for example do not correspond to the most frequent experiences
of life—love and death make for more compelling literature than meal eating and
hand washing. And, of course, the most enduring and dearly held cultural practices
pertain to matters that are central to our view of the world, not necessarily those that
we confront most frequently. Unfortunately, however, discussion of these questions
will have to be left for another time.

References

Biber, Douglas & Randi Reppen. 2002. What does frequency have to do with grammar teaching?
Studies in Second Language Acquisition 24, 199-208.
Brown, Roger. 1973. A First Language: The Early Stages. Cambridge, MA: Harvard University
Press.
 15

Bybee, Joan & J. McClelland. 2005. Alternatives to the combinatorial paradigm of linguistic
theory based on domain general principles of human cognition. Linguistic Review 22, 381-
410.
Bybee, Joan & Paul Harper. 2001. Frequency and the emergence of linguistic structure.
Amsterdam: Benjamins.
Caplan, David & Gloria Waters. 2002. Working memory and connectionist models of parsing: A
reply to MacDonald and Christiansen (2002). Psychological Review 109.1, 66–74.
Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton.
Chien, Yu-Chin & Kenneth Wexler. 1990. Children’s knowledge of locality conditions in binding
as evidence for the modularity of syntax and pragmatics. Language Acquisition 1, 225–295.
Cole, Peter & Gabriella Hermon. 2005. The typology of Malay reflexives. Lingua 115, 627-44.
Conroy, Anastasia, Eri Takahashi, Jeff Lidz & Colin Phillips. 2006. Equal treatment for all
antecedents: How children succeed with Principle B. Manuscript, Department of Linguistics,
University of Maryland.
Cook, Vivian. 1990. Timed comprehension of binding in advanced L2 learners of English.
Language Learning 40, 557-99.
Croft, William. 2003. Typology and universals. Second edition. Cambridge, UK: Cambridge
University Press.
Dick, F., Bates, E., Wulfeck, B., Aydelott Utman, J., Dronkers, N., Gernsbacher, M. 2001.
Language deficits, localization, and grammar: Evidence for a distributive model of language
breakdown in aphasic patients and neurologically intact individuals. Psychological Review,
108, 759-88.
Diessel, H. 2004. The acquisition of complex sentences. Cambridge, UK: Cambridge University
Press.
Ellis, Nick. 2006a. Language acquisition as rational contingency learning. Applied Linguistics,
27, 1-24.
Ellis, Nick. 2006b. Selective attention and transfer phenomena in L2 acquisition: Contingency,
cue competition, salience, interference, overshadowing, blocking, and perceptual learning.
Applied Linguistics 27, 164-94.
Ferreira, Fernanda. 2003. The misinterpretation of noncanonical sentences. Cognitive Psychology
47, 164-203.
Finer, Daniel & Ellen Broselow. 1986. Second language acquisition of reflexive binding. NELS
16, 154-68.
Gibson, Edward. (1998). Linguistic complexity: Locality of syntactic dependencies. Cognition 68,
1-76.
Goldshneider, Jennifer & Robert DeKeyser. 2005. Explaining the ‘natural order of L2 morpheme
acquisition’ in English: A meta-analysis of multiple determinants. Language Learning 55,
Supplement 1, 26-77.
Grodzinsky, Yosef. 1989. Agrammatic comprehension of relative clauses. Brain and Language 37,
430–499.
Hamilton, Robert. 1994. Is implicational generalization unidirectional and maximal? Evidence
from relativization instruction in a second language. Language Learning 44, 123–157.
Haspelmath, Martin. To appear. Frequency vs. iconicity in explaining grammatical systems. To
appear in Cognitive Linguistics 18.
Hay, Jennifer. 2003. Causes and consequences of word structure. Routledge.
Herron, D. & E. Bates. 1997. Sentential and acoustic factors in the recognition of open- and
closed-class words. Journal of Memory and Language 37, 217-39.
Ingham, R. 1993/1994. Input and learnability: Direct-object omissibility in English, Language
Acquisition 3, 95-120.
Ionin, Tania, Maria Luisa Zubizarreta & Salvador Maldonado. 2007. Sources of linguistic
knowledge in the second language acquisition of English articles. To appear in Lingua.
 16

Jurafsky, Daniel, Alan Bell, Michellee Gregory, & William Raymond. 2000. Probabilistic
relations between words: Evidence from reduction in lexical production. In J. Bybee & P.
Hopper (eds.), Frequency and the emergence of linguistic structure. John Benjamins.
Keenan, Edward & Bernard Comrie. 1977. Noun phrase accessibility and Universal Grammar.
Linguistic Inquiry 8, 63–100.
Krug, Manfred. 1998. String frequency: A cognitive motivating factor in coalescence, language
processing and linguistic change. Journal of English Linguistics 26, 286–320.
Li, Hsieh, Laurence Leonard, & Lori Swanson. 1999. Some differences between English plural
noun inflections and third singular verb inflections in the input: The contributions of
frequency, sentence position and duration. Journal of Child Language 26, 531-43.
MacWhinney, Brian. 1987. The competition model. In B. MacWhinney (ed.), Mechanisms of
language acquisition, 249-308. Mahwah, NJ: Erlbaum.
MacWhinney, Brian, C. Pleh, & Elizabeth Bates. 1985. The development of sentence
interpretation in Hungarian. Cognitive Psychology 17.2, 178-208.
Maratsos, Michael. 2000. More overregularizations after all: new data and discussion on Marcus,
Pinker, Ullman, Hollander, Rosen & Xu. Journal of Child Language 27, 183-212.
O’Grady, W. (1997). Syntactic development. Chicago: University of Chicago Press.
O’Grady, W. (2005). Syntactic carpentry: An emergentist approach to syntax. Mahwah, NJ:
Erlbaum.
O'Grady, W. & M. Lee. 2005. A mapping theory of agrammatic comprehension deficits. Brain and
Language 92, 91-100.
O’Grady, W., Y. Yamashita & S. Lee 2005a. A note on canonical word order. Applied
Linguistics 26, 453-58.
O’Grady, W., Y. Yamashita & S. Lee. 2005b. Word order in second language acquisition. In B.E.
Wong (ed.), Second Language Acquisition: Selected Readings, 64-76. Petaling Jaya, Malaysia:
Sasbadi.
Piñango, M., Burkhardt, P., Brun, D. & Avrutin, S. (2001). The psychological reality of the
syntax-discourse interface: The case of pronominals. Paper presented at the workshop ‘From
sentence processing to discourse interpretation: Crossing the borders,’ Utrecht University.
Pinker, Steven. 1999. Words and rules. New York: Basic Books.
Roth, Froma. 1984. Accelerating language learning in young children. Journal of Child Language 11,
89-107.
Sekerina, I., Stromswold, K. & Hestvik, A. (2004). How do adults and children process
referentially ambiguous pronouns? Journal of Child Language 31, 123–52.
Theakston, A., E. Lieven, J. Pine & C. Rowland. 2001. The role of performance limitations in
the acquisition of verb-argument structures: An alternative account. Journal of Child
Language 28, 127-52.
Townsend, David & Thomas Bever. 2001. Sentence comprehension: The integration of habits and
rules. Cambridge, MA: MIT Press.
Wanner, E., & Maratsos, M. (1978). An ATN approach to comprehension. In M. Halle, J.
Bresnan, & G. Miller (Eds.), Linguistic theory and psychological reality, 119–61. Cambridge,
MA: MIT Press.
White, L. 1998. Second language acquisition and Binding Principle B: child/adult differences.
Second Language Research 14, 425-39.
Wong, Bee Eng. 2002. Acquisition of English relative clauses by Malay speakers. Journal of Pan-
Pacific Association of Applied Linguistics 6, 61-77.